Це відео не доступне.
Перепрошуємо.
Dead server fan autopsy. (Unusual circumstances.)
Вставка
- Опубліковано 2 лис 2016
- An unusual autopsy on a dead server fan that was killed by the placement of a huge magnetic field outside the case it was in.
This fan was sent to me (packed carefully in chocolate coated marzipan) by two Danish UA-camrs who had inadvertently caused it's demise in a collaboration they had done together involving some servers and the biggest neodymium iron boron magnet they could find. You can see the original video here:- • Super Neodymium Magnet...
If you enjoy these videos you can help support the channel with a dollar for coffee, cookies and random gadgets for disassembly at:-
www.bigclive.co...
I came here for a fan autopsy but what I got was fear of magnets.
Don't fear them, just respect the big magnets and know what they can do. They can be found in industrial equipment and the braking system of rollercoasters.
bigclivedotcom mayby the big magnet atractet the rotor and rotor got stuck and the coils shorted the diodes out?
bigclivedotcom
Massive inductive load death.
The redstoneblock computer fans will stop if they're blocked.
Phos9 o Yes that is true i forgot that
My horrible(ly) dangerous/funny magnet story. Growing up around age 10-12ish we had access to two of these death-magnets. His father was VERY insistent we never touch them or even bring the two in the same room together. Well we were dumb stupid kids and of course the first thing we did when alone was to pull the magnet out. We took the magnet and set it underneath the glass living room table (yes we were stupid kids). We then proceeded to make a game of "skipping" thin metal washers over the table and watching them ricochet in different directions as they came in contact with the magnetic field. Which honestly even today was pretty cool effect. Once again being the stupid kids we where we refused to allow my friends brother to play with us and the magnet. Oh course he goes off and gets the SECOND magnet on the other side of the house and bring it over to us so HE could play with one too. Everything happened so fast one moment we where laughing and tossing washers, the next second the magnets, washers, and entire glass table EXPLODE. My friends brother got too close holding the second magnet, Magnet 1 on the floor jumped up to join with Magnet 2, Magnet 1 took the glass table for the ride and thankfully magnet 2 got ripped out of my friends bothers hands. When they came together the huge glass table just exploded into millions of beads of glass. Thankfully the only injuries where to the brother which had minor cuts from the jagged magnet getting ripped out of his hands, and then later our asses from the beating we got. The magnets came out of some military equipment, I'm thinking a radar system but might be wrong. Those two magnets are STILL (25 years later) at my friends house, permanently bonded together with a thin glass layer in between.
Great story! Thanks for sharing!
have you ever considered finding some way of separating them
@@jemzomaclain That would be incredible danagerous to separate them
@@aldrinw706 It's not dangerous if you know what you are doing. The biggest question is how big those magnets are and what is the distance to keep them apart. Basically you'd have to build wooden rig that starts with kind of wedge and then push it in between. Everything else is just matter of proper design to avoid it accidentally jumping out and hitting each other again. If the wedge is wide enough, has kind of rails on the sides and possibly has steps that prevent magnets from sliding back, then you can do that. The bigger question is - do you want to? What will you do with them afterwards? You will always have to keep in mind where you have them and never let them meet each other again. Keeping them in one place is just easier so unless you have a reason to separate them, it's pointless.
^ found the engineer
Having watched this I did a small experiment on my breadboard. Apparently strong magnetic fields will cause MOSFETs to go into an excited state. The neodymium magnets i have are nowhere near the strength of the ones used near that fan. The MOSFETs I used were the NPN TO220 variety and I was only running them off a USB power supply. I pulled the gate to ground with a 100 ohm resistor and applied the field to the body of the TO220. To my shock, it went into full conductance. Enough to run a small DC motor at least. Those servers have a high amperage bus for their fans because of how hard they have to turn those fans to move that much air in a tight space. I'd imagine what happened is that all the MOSFETs went into full conductance in the presence of that huge magnet turning everything on at the same time. When I took the field away it went back into normal operation.
The magnets I used are one inch in diameter and about a half inch thick. They are considerably strong and I don't handle them with my hands directly between them. I've already seen what they do to a steak to show my son why he shouldn't take them to school. To get them apart I have to use an aluminum shim and some special handles I had to make with 3D printed parts and some nylon strapping. I'm not strong enough to pull them apart with my bare hands. There simply is no way to grip them well enough. I can't even imagine how strong the one that those guys used on that server was. Scary stuff...
Nice.
fwiw aquick search of google scholar says you can get Hall effects in commercial MOSFETs
Well, THAT's interesting. But that still doesn't explain blown zeners - unless they are below the 12V supply voltage and the current limiter went wide open. But, then they'd be conducting at times that doesn't make sense.
I'm now wondering about the possibility of the power supply going wild. In the past, when I've torn them apart, the 5V is regulated but the other voltages are not. What else in the server is powered by 12V?
ExStatic Bass well, make a youtube video about it ! Won't believe it 'till I see it
Another possibility with the magnet is a straight up mechanical failure of the semiconducting oxide layers in the transistor or zeners
ExStatic Bass maybe the magnets messed with the PSU switching FETs and sent high voltage on the 12V rail?
Great video! Thanks for taking your time to do an autopsy on the fan. The result is surprising. I thought it would have been some form of mechanical failure - maybe some insulation getting scraped of the magnet wire in the coils causing a short circuit. I didn't expect to see burned/exploded electronic components but it explains the weird smell coming from the server :)
Usually electronics cope quite well near the magnet. I guess the magnet must have triggered some unusual event in the server causing a spike inside three of the four fans in the server. Odd but interesting.
I remember seeing your video a few weeks ago! It was awesome. Subscribed.
Moving a Monster Magnet across the stater coils turns the fan motor into a generator. The stronger the magnet and faster it moves the more voltage your going to induce in the stater coils. With a pulse of high voltage the zener diodes failed to a become a short and protected the mosfets. Just my two cents.
I disagree. I think it was the hall effect sensor sending weird signals out of phase to the controller. When you wan't to regenerate some of the power you put in to a BLDC motor, you pulse the coils out of phase resulting in high voltage spikes recuperating some of the mechanical energy back as electricity. Since the circuit was not designed to handle regenerative braking of the motor, the high voltage spikes killed the zenners. The 6 pin package ic is the PWM modulation controller. After the zenners blew it had to handle too much current and just puffed away.
Eviltech I'd imagine you're the closest out of everyone to what really happened here. that strong of a magnetic field probably totally saturated the hall sensor and like you said, the resultant driving of the coils out of phase caused quite a lot of voltage. I have a couple of these fans in a closet somewhere, I'm tempted to hook the hall sensor as well as the coils up to an oscilloscope and find the biggest magnet I can to reproduce this
Jon Sands You do't need a big magnet, i have killed fans simply by putting my phone's magnetic flap next to the hall effect sensor, but in a different manner, because those were cheap simple fans that did not care if the field was spinning or not, they just powered the coils without any additional logic.
anybody else think Clive should go into narrating audio books?
Yes please!
I'd buy it even if it was 50 shades of grey
@thorerik what about fifty thousand shades of grey?
Make him narrate 'The Art of Electronics (3rd Edition)'. c;
256 shades of grey at the very least. But i prefer 16777216 colours.
I like when you have printouts of the circuits you're analyzing. It's simpler to comprehend.
Well he does try to reverse is your even the simplest thing sometimes but yes I do agree and it helps me even even though I'm quite into the stuff I'm sure everyone else as well so keep it up Big 5 please please please please
Oops just realized I said Big 5 instead be quiet actually that wasn't me that was Google Google strikes again LOL
Hi Clive, I have a background in motor control and can perhaps add a little to your explanation (we're FB friends under a different name).
That centre tapped arrangement is called a unipolar drive (the mosfets only drive one end of the winding at a time) as opposed to bipolar drive with the h-bridge. This can only be for cost reasons, there is no control difference between the two methods and unipolar is far less efficient as it only uses half the copper at any one moment.
The zeners are there to protect the mosfets in the case of transient voltages exceeding the mosfet Vds breakdown limit. They must have at least 24V Vz, as when one mosfet switches on, the winding at the other end will go to 2x the supply voltage as the winding acts like an auto transformer.
It is common when driving inductive loads from a bottom switch like this to put a reverse diode across the load so the load current can safely recirculate through the diode when the switch goes off, preventing a high transient voltage. But in this arrangement, if you do this then when you turn on one mosfet, the opposite diode will conduct and short out the magnetic flux. So you have to use this zener arrangement, or something similar.
What seems most likely to me was that a change in the magnetic flux through the windings, caused by moving such a powerful magnet in their vicinity, induced a voltage in the windings sufficient to exceed the zeners' capability and they failed short. Were the fans running at the time? If so, then the zeners would already be operating near their limit and it would need only a few extra volts induced to kill them.
The transistor at the top could well be driven by the PWM input signal, with the low cost driver IC just providing commutation. This would be a very cost effective solution, and would be easy to confirm if the PWM wire was connected to one or more of the resistors next to the dead part. Of course, once either of the zeners failed short then this transistor would have the 12V supply directly across it and would fail spectacularly.
Moral of the story? Keep BBMs (Big Bastard Magnets) away from electronics, and especially brushless motors!
I think you may be on to something. My thoughts were that the diodes were there acting as shunts.
My theory is that when the big magnet came close, it overpowered the hall effect sensor. When that happened, it probably caused the switching system to not realise that the stator needed to be switched and caused some feedback in the coils themselves, which then caused the over voltage through them. I could be quite wrong, but without being able to actually reproduce and test as it going on, I don't know for sure.
I believe it's called a cash diode or a freewheeling diode sometimes.
But also Audrey Lake Wales solenoids and alike there is a reversed. I owed surges from blowing driver transistors or other driver components.
Also in Electro pneumatic pipe organs there's the diodes across relay and other magnetic coils to prove checked the switch contacts and also of course on a solid state logic driving any part of the pipe organ.
I had to replace a few diodes here and there on the old Orcutt that went short.
And actually wind blowing a fuse but couldn't figure out what caused it first and Tom like a diode gone short.
Remember diodes have jumper Envy.
You know how sometimes on a laboratory power supply or the power supply is protected before diode in Reverse across the input leads to prevent damage in case of extra extra polarity reversal or feeding power end the wrong way.
That usually will fried a diode if it's bad enough.
Also same thing often happens in crowbar circuits for reverse polarity protection the diode usually whines up failing short protecting the device.
And oftentimes if a scr track is used same thing it will fail catastrophically. After it blows either a fuse or source to power supply.
I've repaired a few devices that had inline fuse or fuse somewhere that the polarity is reversed in a diode cost of a fuse blow because it was forward biased by reverse current Alpha polarity reversal event which is why it's supposed to do and then they said it just kept blowing fuses and found out that diode it was there against player Tavor so go ahead field short because of someone not me reversing the polarity trying to hook it up. Edit it's called a catch diode. A cash diode although it could turn into a cash diode if you ever replace it. Is introllsive replace the device buy buy a new one.
*googles 'magnetic defense gloves*
Not sure what I expected.
I can totally imagine some gloves reinforced with sections of carbon fiber tubes that won't let your fingers be squashed... :)
+Lucas Nice pic dude :p
All gloves will save you from losing the tip of your finger, as the gloves will act like a tube of toothpaste, pushing your mangled finger out from between the magnet and metal so it can get closer. Unless they are fingerless gloves. That would be like wearing a headless condom.
@Pass the Spider and you would be the one say about Googling it ereser image checks out LOL.
Leaving some of the electronics aside, I have to admit I do love the shape of those fan blades. They look very efficient. I reckon if you double the blades, double the poles and modify the circuit for x4 current, you can create a quiet, forceful and yet controllable component fan.
A magnet with that much flux could easily induce a massive current even without a fast motion. I suspect this is what happened.
rjk you still need a fair amount of movement for this even with a strong magnet, if you watch the server video they could just barely drag it over I do not think this is the case
i agree, with rjk. Also, I think that the motor windings might help to induce high enough voltage to damage the components - that's why the Zener diodes failed - and that current regulator - which were in the path of the current. But I'm not an electrical engineer, so this is just a guess :) .
This was my first thought also but it's a long time since I did any theoretical stuff to do with flux fields etc which could be used to determine the induced voltage. Even a brief spike could be enough to break down a zener (assuming it was high enough) I would have imagined. Somebody out there knows the answer!
well if you disconnect a spinning low voltage permanent magnet motor it'll generate a surprisingly high voltage across it's terminals until it's stationary, possibly giving the person disconnecting it an electric shock. but in this motor the stator has the windings so I don't think the magnetic field from the super magnet could have been moving rapidly with reference to the copper. I'd have thought the permanent ring magnet on the rotor would have shielded the windings from an external field if anything (rather like the iron sheets in the magnet packaging video made by braniac75.
I don't think we'd know the answer for sure without carrying out tests of the best (destructive) kind
if the fan were spinning, that count as a fast moving object over the magnet point of referential, don't they?
What a great surprise to see that video a month ago only to see Clive autopsying the fan that was damaged. Never would've expected this! :)
I was suggested to send the fan to to Clive,, and it seemed a awesome idear.
Do you watch his vids too?
I kinda wanna see the slowmo guys let two of those magnets meet.
Braeden Hamson yes!
Already done
@@maherjrad7951 seriously I would like to see that one for sure what next two trains in a cornfield meet at full speed fully loaded 100 + tons each AKA head-on collision.
I've got a bunch of neodymium magnets, they're disc shaped 20mm x 10mm........ When I first got them I put one on top of my hand, another underneath, which was great fun until they flipped and pinched the skin between thumb and forefinger, it took several minutes to part them using a variety of tools and they left a nasty bloody mark for almost 2 months.
They spent 12 months stuck to our exercise bike because my GF couldn't get them off the frame and now they live in my garage a long way from anything magnetic.
DO NOT f**k with neodymium magnets (unless they're the stupidly small ones) :p
Well, laddy, you'll nay be doing that again, will you.
Even the small ones can be dangerous if two are eaten (as they sometimes are by children).
I have a list of politicians who'd be improved by more magnetism in their personality, at least until I can get some Polonium...
What I'm wondering is how do they even ship magnets like that? Or maybe you have to pick them up yourself? I can only imagine the complete chaos of the magnet interacting with metal things in other packages, as well as the walls of the delivery truck!
The RPG Chick they use field dampers (chunks of iron) that drag the magnetic field in close preventing problems.
That was surprising. I was guessing the magnet in the fan part lost it's polarisation or got a new one in a disadvantageous angle.
magnets so powerful that they have their own bureaucracy surrounding them
Core saturation! The magnetic field was so strong that it saturated the core of the stator, causing the winding to stop to be an inductor and become a wire!
But how would that fry the Zener diodes?
Jack White maybe that by reducing the inductance, it cause more current to flow, so the magnetic field collapse faster after and dump more energy faster? but really, not sure. I think the saturation is only part of the equasion...
Lower inductance would mean less energy dumped from the collapsing magnetic field. "Speed" does not matter here, I would think, because "speed" is determined by the other components around the inductor and they didn't change. But a saturated core could mean unrestricted current flow through the mosfets, so I would have suspected the mosfets but not the Zeners, with core saturation.
Jack White A lower inductance cause a more rapid raise in current...
The zeners would start conducting the moment the inducted voltage exceeds the voltage they are designed to conduct at. if the current is high enough, the result is a dead diode.
My guess is simply saturation of the motor coils cause by the strong magnetic field. This would greatly reduce the effective inductance of the coils and allowed excess current to flow.
en.wikipedia.org/wiki/Saturation_(magnetic)
Saturation, yes, but of the stator IRON, not the copper coils. The big external magnet simply swamped out the piddly rubber magnet inside the rotor, then proceeded to saturate the laminated iron stator. Once saturated, only the quite low resistance of the coils would remain, instead of their (usually high) reactance. It's like the coils became shorts, drew massive currents, and poor little semiconductors got killed by the angry pixies. Plus the rubber magnets probably got re-magnetized with improper poles.
Makes sense.
YodaWhat This does not explain the burned-out zeners however.... In contrary !
Dead conversation, but for the benefit of anyone else reading it, I think this is a plausible explanation. The inductance gets cut by fivefold, the current goes up five times, and nowthe zener diode has to deal with the same duration of current pulse, with the same zener clamping voltage but fve time the current and energy. I think the motor was stalled/abnormally magnetized for many seconds to actually scorch components like that.
This is awesome, because about a year ago, about the same time you posted this, I had to repair a pair of server fans. The fans themselves were the same model, scrapped from another server, but had an entirely different set of wires coming out of them. So I had to rip them apart, desolder the wires, and swap the wiring around.
I'd just like to say thanks for your videos. I couldn't stand the bland way school taught us about electricity and circuitry (nor could most of my class from what I could tell). However, your videos are highly educational whilst staying entertaining and strangely addictive! So thank you and I think I speak for many of the younger viewers when I say please, don't stop making videos!
Im fairly certain that overvoltage was the problem. Those zeners are obviously for overvoltage protection from the inductance but werent made for the extreme voltage that magnet must have induced. Change in Magnetic field is what causes a voltage to appear on those windings and the ridiculius intensity of the field compensates for the low velocity. Those guys should try moving one of those coils around the magnet, with a meter across it.
The magnetic field confused the hall effect sensor, both FETs opened, possibly a condition the controller, wired in non standard configuration caused, it definitely allowed the condition to hold, until the weakest link broke
The magnets in the rotor will have acted as conductors moving in a (very strong) magnetic field at 3000 rpm or whatever - acted as an alternator and arced over or induced large eddies in the stator and pcb?
Those magnets in the rotor (as in, the outer cup shaped part that actually moves) are not electrical at all. They are made from a solid piece of strongly magnetic ceramic. I'd doubt they'd arc if they were spinning at 3000 rps directly under the magnet.
The wound copper coils in the center of the motor don't spin at all.
I am a manufacturing engineer at a major fan company.
1. That design fan is called an inverted fan. Its called that because the magnets go on the outside of the stator instead of inside it like on a conventional motor. Very common for small fans because it reduces footprint size.
2.It uses a 2 phase setup with the center tap being pulled to VCC. Benefits of this is allows for very simple circuity. As each phase is pulled down by the FET, it fires that side of the stator.
3. Those zeners are to protect the FETs from the back EMF of the motor.
4. When the fan has a small failure, it tens to cascade as high current draw and back EMF creates lots of heat and burns up everything.
It's probable that that bit that melted was a current sensor of some sort. A lot of fans, if you stop them with your hand, will stop trying to spin after a couple seconds and then try again after a few more seconds, probably to avoid burning out the motor.
WTF, Are all the youtubers i subscribed to friends or what? Pretty awesome though.
Hatagashira I was thinking the same thing
Hi Hatagashira
No but we all work hard, on keeping you happy, and entertained :-)
Illuminati confirmed!
My PlayHouse Mission accomplished! :D
Thanx, Hatagashira
Maybe the magnet saturated the stator core and it no longer acts as an inductor, but as a dead-short circuit
You are right about the PWM, the specs say: if unconnected it must go to 100%. Some also will go full blast if you go bellow the minimum, which is usually around 30%. PWM is 20kHz 5V
Maybe the neodymium magnet caused the stator windings to _produce_ more voltage than the Zeners were rated for, blew them out (shorted), and this resulted in TR2 burning out.
I think that's been covered but possibly yes.
If the magnet caused the stator windings to cancel their inductive reactance, then assuming also a negligible resistance, too much current will flow killing both zeners and the regulator. In the alternative, the moving magnetic field may have induced a large field in the windings to blow the damaged components (induced polarity unknown). I watched their entire video; the servers were running with the fans spinning.
I was expecting the motor to have a weakened magnetic field from being exposed to an "uncontrolled" external field
The magnet shown in the picture is "Der Todesmagnet" (literally "the death magnet" in German). This is the magnet Barry Wels used on stage when he demonstrated the flaw in the Wink Haus BlueChip lock.
Such a massive magnet would have introduced a large current in the coils as it moved by, those zener diodes are effectively connected across the windings, so could have been damaged by that.
Potentially?
I once connected the minus pole to the signal wire on a PC fan and found out it worked. Although at about half speed. This level of electical details blows my fan-riddled mind. Awesome stuff!
The Zener diodes may have been used as a clamp or snubber for the voltage spikes when the mosfets are turned on and off. Zener diode do have power rating and will burn up if too much current flows thru them while they conduct.
The standard data sheet schematic looks like a "H-bridge" circuit which allows the motor to be reversed. Since cooling fans spin in one direction, only half of the H bridge is needed.
Is there a 'Forensic Electrician' job? Because that's the vibe I got from this video. Love it!
Ooh my favourite youtuber dissecting parts supplied other favourite youtuber. What a small world. You really are great Clive, please don't ever stop making videos. I have genuinely found you an inspiration. Nice beard too!
New kind of food chain :-) FAN mail
Morten, you are fantastic. Keep up the good work, 15000 people agree with me on that!
ben howard
Thank You!
The way the fans were mounted in the chassis, here's a fairly mechanical idea: As the Big Bad Magnet™ was put on top, the fans started to alternate (*badumtishhh*) between almost blocking and generator mode as the magnets on the rotor were rotating through the external field.
I wonder if the magnet was so strong it locked the rotor and the stall current burned out the components. That's my guess anyway.
Wise and useful words about the dangers of neodymium magnets, thanks Clive.
I've suffered many strange and unexpected injuries; from brittle metals or ceramics shattering.
Eye protection is something to take very seriously too, as I have been been literally _Shot_ by bits of neodymium and Ceramic in the past
I cannot get rid of the mental image of a guy sandwiched between two magnets and stuck to an ambulance.
We live in a bizarre cyberpunk dystopia. Neon lighting by Big Clive Dot Com...
Very nice detail printouts! If I were Clive I would collect printouts for every component inspected and make a wall collage :D
Actually he should turn it into wallpaper for a shop and it's all else by now LOL
My guess is that the magnet caused the fan's motor core to saturate. When that happens, the motor windings act more like a short circuit rather than a inductor. The excess current likely caused the electronics failure.
How would shorting anything out in that circuit conceivably blow the diodes?
Whatever inductance was still left over would now be "charged" with wayyy to much current. Or, the high current burned something else (mosfets? Chip?) and that led to an abnormal voltage or current in the diodes.
From your description, I think that the big magnet saturated the stator core, so when the transistors switched on, the current ramped up *much* more quickly than the controller could handle. (It may have a minimum "on" time for the mosfets. This could cause the zeners to see far more current than they were rated for.
I know this is a year old. But my layman's diagnosis? Perhaps as the magnet was passed over the fans, the blades slowed because of the magnetic field. The fans sensed the slowing blades and tried to speed up the blades only to denied by the magnet causing a overload which blew out the components.
Maybe the magnet saturated the silicon steel the winding is on, so that the winding dropped considerably in inductance and essentially behaved like a short with a very low-winding resistance - which then could trigger the follow-up failures.
That sounds plausible.
My first thought (before you checked components) was that the magnetic fields of the rotor magnets was altered by the enormously stronger neodymium magnet, giving the sensors conniptions.
The way i see it is that the Magnetic field change the domains in the iron core. hereby the motores goes into saturation,
and it is left to the copper resistance to restrict the current flow.
HI Peter,,, It is on EEVblog to day.
I just found out recently you can separate that double stick they use on those with quick dry electric spray. Weakens the industrial adhesive to the point where it feels like cheap dollar store double stick.
That magnet was colossal. Even moving slowly, a field that intense might have generated a significant voltage. Was the fan running? if it was, maybe the big magnetic field coupled with the magnets in the rotor, temporarily making the little magnets substantially stronger, which might have caused feedback through the windings?
I wonder if, perhaps, the big magnet re-magnetized the rotor magnet in the wrong direction, stopping it from working properly.
I saw the original video where the fans & HDD's failed. My first guess as to why the fans failed is that a switch mode regulator feeding them became severely de-tuned by the added magnetic field, resulting in a much higher voltage than normal being output to the fans.
Thing about kid's hand getting smashed was interesting... i am horrified
The magnet can lift 1200 kg,, it is not all fun :-/
My PlayHouse Well lifting thing that weights1200kg isnt that scary... i mean not always... not when there is a human, or animal or something between magnet and this 1200kg thing
If you have a big magnet near a brushless motor, it can overwhelm the power control circuit, which would make sense as to why that component exploded. The amount of current it drew shot way up beyond it's capability. I'm surprised the bearings didn't go, then again they are heavy duty high speed bearings.
What probably happened is when the magnet was passed over it while it was running, probably causing the rotor to stall and stop. Now normally the hall efffect sensor would catch this however iff it was too also affected by the strong magnetic field, it probably didn't work as it should so instead of stoppingor lowering the flow of current to the coils. (what typically happens if one stalls.) it started pushing large amount of current til something go pops. Not done this exact thing with a large magnet but I have done stall test on different variants of brushless DC and some faliure mode ended up pretty similar to what im seeing here on this pcb.
Another theory is that the hall efffect sensor itself alone was th culprit. ITs what the drive circuit relis on to drive the coils. If the hall effect sensory gets a bad reading from that big magnet, it could have caused the brushless dc motor to miss drive. (pulse the wrong coil at the wrong time) and you end up with an inductive spike in the coils.
Ive seen bldc motors fail in many ways. Its not always easy to determine but based off what I see here, I think one of 2 things happened. The rotor was stalled or even slowed by the magnet, causing the coils to be misdriven because hall efffect sensory couldn't correctly detect rotor position. Switched wrong coil, inductive spike and pop.
MY 2nd idea. Hall afffect sensory was affected in a way that instead of switching, one coil managed to stay on, stalling rotor, causing higher current flow, avalanche affect.
I have always been very interested in magnets, as a boy I used to visit the dump grounds here in town, back then they were not manned, and anyone could sort through the trash in the pit, or visit the automotive graveyard that had cars stacked 4 high in a huge maze (before car crushers you know). A buddy and I used to hit the dump and dig into those old wrecks, I would take the speedometers out and get the magnets out of them. The magnets were great fun if suspended by a thread and spun, each would spin the next for a dazzling display of magnetic fun. Who would have thought, back then (early 1960's) that we would have such powerful magnets now! Thanks for the continuing education. I always have believed that if I continue to learn one thing new every day that I would not age quite so quickly. Thus far it must be working, as I still don't feel all that old at 65.
OH OH I REMEMBER WATCHING THAT SERVER VIDEO
I remember making it :-) did you like it?
+My PlayHouse
I know that I remember that vid. It was quite interesting to see.
HI theLuigiFan0007
Humm "interesting" is not always good :-)
My PlayHouse
Indeed. Depends on the perspective. ;P
I honestly expected more immediate drive failures. Guess they're a little more durable then I though. Still, it goes to show, big magnets + servers = bad. XD
Reminds me of a Defcon video of how to quickly destroy a server to avoid it's contents "falling into the wrong hands". As I recall, thermite was tried, unsuccessfully.
My thoughts... and I don't know if this has been touched on, but hey. I suspected this when I first saw the failure in the original video - the control chip got confused by the hall effect and shot off both windings at once! Then, that caused it to short-circuit, weld something together, throw full current at a coil, and of course, cascading failure by overheating (when the fan stopped turning, as it quickly would when a single pole is being held active, it'll apply forceful a braking action to the fan).
Of course, that may not have originated with the hall sensor, but rather (as ExStatic Bass mentioned) started with the MOSFET switching both gates to activated -- causing a short, since the motor - I think - works with switching polarity. Only one (or two, in the case of 4 driving 2 linked sets of coils) of those gates should ever, ever be open at a time. If two of opposite polarity were to be open, it'd immediately cause a short circuit (thus the principle of an "H-bridge").
On the right track, perhaps?
The easiest way to remove those magnets is heat: Hold a blowtorch on them for a couple seconds and they'll come apart. It may get warm but it won't hurt as much as cutting the hand off around it, or so I think.
In any coil, the voltage is proportional to the change in magnetic field with respect to time, so that makes sense.
I would guess that the reason for the high voltage across the diodes would have been because of the magnet and the EMF generated by it. Because those pins are connected directly to a coil, the magnet would have produced an electric field (due to Faraday's law). Also, note the servers were running at the time.
Exposing switch mode supplies to strong magnetic fields while in operation can make them very angry.
Although the larger magnet was moved slowly, the fan surely was rotating inside the field of the much larger magnets field at something like 2000 rippems... thus created the same effect as the larger magnet moving faster past something more static... Surely in one pole orientation the fan would have had a super sonic kick up the arse as it moved in/out of phase/magnetic orientation and became a generator for every 1/4 cycle. Now thats the sort of magnet for finding a needle in a heystack.. ziiing.. found it.
@bigclivedotcom maybe the fan was simply blocked due to the enormous magnetic field of the neodymium magnet and then the current through the windings went crazy
I still think that massive magnet sliding into place induced a massive voltage in the windings and electricuted the pcb, On a small magnet you would have to move it fast, but that massive one would produce a lot with slow movement.
In case others stumble across this comment, it's worth noting that the peak field strength is not as important as how fast the field increases. A big magnet will start magnetizing the motor when it is a few feet away, and then gradually increase to a maximum as it gets closer. It would actually have to fly towards the motor faster than a small magnet to create an equal △B/△.
My first thought was the magnet would have saturated the stator core laminations and this could have lead to the blown series transistor, possibly current sensing as you say. But I can't see a mechanism for the zeners getting blown assuming they're >12V.
If magnets this big collide raw, it will release more(and faster) shrapnel than a very sizeable grenade.
It's a sight I wish to capture on my high-speed camera, but each of those magnets costs way too much, and you'd need two(or more if you fail).
Maybe it is much simpler
Magnet stopped rotor, so rotor's magnet became static, so no change in magnetic field from magnet side.
Winding is simple choke and works like come in absence of external source of magnetic field. Controller have switched off wrongdoing, but due to nothing taking this energy, magnetic field have collapsed and got converted to very high voltage, which in effect shorted diodes.
That's about what I think happened, but I'm not positive.
If i had to guess, i'd say the magnetic field stalled the fan, and power was going into the fan but had no outlet so it just pooled up (like a giant inductor) and burned up the electronics.
I reckon the magnetic field of the neodymium magnet saturated the stator core, allowing much more current to flow than it would otherwise have. When the mosfet closes a lot more current goes through the zener, burning it out.
Looks like Morten's writing on the fan handle didn't survive shipping. He wrote "BAD" in knockoff Sharpie on the orange hotswap handle.
HeHe Yes it was still there.
btw, to separate two strong magnets you can slide them off each other. another thing, you don't need two of these magnets to crush your hands,
while these magnets require proper handling (there is actually a license requirement). they are not as evil as you make them. like any other tool wrong handling can cause harm
Golly Clive, never in my darkest dreams did I consider that something children use and play with can at the far end of the line be so damn dangerous. Another eye opener and a scare as well albeit I will never have one!
I ran a neodymium magnet across a running hard disk. All that happened was the disk stalled but after a reboot there were no defects to the disk surface.
I did not expect a connection between you and Morten (MyPlayHouse). nice to see more about what actually happened there.
Well I was suggested to send a fan (fanmail) to Big Clive.
And the marzipan it came packed in was much appreciated.
bigclivedotcom Well good packing material is very important :-) Thank you for making such an awesome video. I can see that over 1800 has been by me watching the video on killing the fan. I did a video last night telling my subscribers to go watch your video,, I think a lot of them did.
there's also another possibility of messing with the magnet on the fan and jamming it in place and bad things happen when fans are jammed plus the electrical force (yes magnets do induce EMF) could have stuck the mosfets wide open and of course the magnetic sensor could have been fooled by that huge neo magnet, tricking it into ... you know the rest
Clive you should do a meet and greet when you're in Glasgow!
Maybe the motor locked mechanically from the magnetic forces and the control circuit keep pulsing the windings until the repeated fly back pulses burnt out the zenner diodes. The magnetic field from the magnet may have also saturated the windings which would have made things worse.
Wow! My first thought was shoot-through, two mosfets kicking on at the same time and connecting positive directly to ground. Doesn't look like that's how this one was laid out, though, and the mosfet itself isn't blown up. Gotta just be induction from the big-ass magnet passing over those coils.
Think for a moment. If you introduce a magnet to a rotating magnetic field, you get electricity. Its how the fan works in the first place. Anyway, the super magnet created a back current. Apparently strong enough to blow the diodes and continue until it met the input current at the transistor.
The fan was RUNNING! when the magnet was brought into its area, so it could indeed induce quite a lot of current!
The part that was running was a spinning magnet. Not a spinning coil.
I think what might have happened to the fan may have been a current surge caused by the combination of the collapsing magnetic field, reinforced by the additive "static" field from the permanent magnet.
11:28 - "...and the other magnet gets attracted to the magnet you're holding..." then you wind up with little baby magnets that you have to support until they're 18 yrs old. Not to mention the awkwardness of having to meet the owner of the other magnet. LOL
Huh. When this started, I would have guessed they demagnetized the magnet in the fan blade. But I think what happened is moving the magnet induced a high enough _current_ to blow the zeners. Remember - a zener is a diode. It just happens to have a controlled reverse breakdown voltage. You put a diode across a coil and move a powerful magnet, it shouldn't be too big a shock if you exceed it's current rating and blow it.
My guess would be that the zeners conducted every time a coil was shut off, but the magnet either held the fan still by brute force or saturated the stator/weakened the rotor field. The chip told the coils to turn on for say, 5msec as per normal, but with lower inductance in the motor the current rose to a higher level. The reduced inductance was offset by the higher current, E=L*(I^2)and the zeners had to dissipate more power--> overheating--> open or short circuit failure. That, or the hall sensor was overwhelmed and the cylinders got fired out of order. Either way, it was probably stalled and heating up for several seconds at least. Wonder if it would be worth finding a similar fan and deliberately burning it by holding a small magnet directly over the motor, while monitoring current consumption or even puting an oscilloscope in it.
The exploded part is most probably the MOSFET used for PWM control. ZD things are either Schottky diodes to protect MOSFET body diodes from reverse currents or actual Zeners to limit voltage across the MOSFETS by sinking DC current through coils in case of overvoltage.
The zener arrangement of having a large zener (e.g. a 24V zener for this circuit) in parallel with the FET is that when the FET turns off, the continuing current in the coil boosts the voltage until the zener lets the current through at 12V *above* the +12V rail. Each coil does this separately, providing two overlapping electromagnetic waveforms which combine to form a rotating magnetic field which rotates the permanent magnet rotor in a direction dictated by the phase relationship between the two control pulses.
Fair enough, could well be this. Still not clear why they didn't just put Schottkys across the coils then. Probably to implement what I said as well with no extra components. If so, that's a pretty clever design here!
great video. I really like the way you pulled in the pictures of the circuit board and explained how it work to the best of your knowledge. Thanks a lot
i have had many accidents with a hand full of 1inch cubed rare earth magnets. luckily i never broke anything. many pinched fingers and needles stabbing me. putting them in different coat pockets worked well for a while until they ripped out of the pockets to smash together. gotta respect those magnetic fields
Sigh. Now I really want to see someone take a bunch of those magnets and build one of those 'perpetual motion' devices with the spinning ring of magnets. I'm sure the construction would have to be planned very carefully and very strong materials used, but I just think it'd be entertaining to see.
My school curriculum included that center tapped coil circuit. I think it's mostly made to save space while having every advantage they need from the h-bridge arrangement.
I guess I'm the only one who expected the failure mode would be the permanent magnet in the rotor getting demagnetized or magnetized in an orthogonal direction.
If the rotor was stopped by the magnet, the circuit thought it wasn't going fast enough therefore giving it more current resulting in failing of other componments
The fan's initial spinning magnet+giant magnetic field could easily have cause a voltage induction that would spike backward through the stator.
EEVBlog got sent one of these too. Have to say, your explanation was more thorough. I know you don't credit the magnet with having been capable of generating a big enough spike while being dragged over the case, but those magnets are truly powerful...
perfect fridge magnet
I hope you're referring to the magnet of the fan and not the mega magnet because they'll be one refrigerator magnet you would never be able to remove so what's the point of refrigerator magnet you can't remove? LOL.
Instead of hanging magnets on the refrigerator, you hang the refrigerator on the magnets.
The neo magnet probably saturated the core laminations of the fan which would cause the windings to appear as a short, blowing the switching devices.
Good theory, except the switching devices weren't blown. The mosfets were fine.
Awesome to see that they sent you the fan to take a look at.
FAN mail!
Your comment came very windy...
Drakkar Calethiel Hope it, did not blow you of your feet :-)
The little fan doesn't use rare earth magnetic does it. also I bought a pack of a hundred little rare earth. magnets ,the little suckers are hard to hold on to.so much power on something that darn small is amazing.
I saw that video before I was subscribed to you from a random suggestion on UA-cam, and here you are, with one of the faulty fans.
UA-cam's algorithms are really interesting.
Overvoltage on the 12V rail is not very likely. The server is usually fed from two power supplies. If so, the system would have turned off or it would have damaged other components in the system like the motherboard.
More likely the the thing that that mosfet got excited through the magnetic field. Or the hall sensor reported a static magnetic field, that was interpreted as a not running fan, leading to increased current throuh the coil /mosfet.
I hve never seen a failure Mode like this. Since that fans are usually roboust to beeing blocked. Btw. 50 mA is not really much for server fans. They can consume up to 2 amps in operation each.
It's nice when all our UA-cam friends come together :)
Hi Knuckles the Echidna
Not much gets by you!! :-)
+Knuckles the Echidna: It certainly is great when they come together to make a video. :D
Creates interesting videos and also gives you an interest in the other's channel if you'v not seen them before.
HI theLuigiFan0007
I am amassed how many of my subscribers was over here before I even told them about it :-) Yes this kind of cooperation works really well,, both for channels and viewers.
My PlayHouse
Indeed it does. I'd already known about your channel and had been subscribed, but I still did end up watching a few more of your videos right afterwords. Guess that's another way it can increase viewing, simply reminding people of a channel again, and renewing their interest. :D
HI theLuigiFan0007
Well we are all fighting for you eyes,,we want them! look at me :-) It is great!